Using oxygen-15 water positron emission tomography to measure regional cerebral blood flow (rCBF), a marker of local neuronal activity, we have found that normal subjects performing tasks involving working memory use a cortical network including dorsolateral prefrontal cortex, inferior parietal lobule, and inferior temporo-occipital cortex. In other tasks related to prefrontal cortex, we have shown, with a maze task, that acquisition but not retrieval of spatial data involves frontal cortex (Van Horn et al, Brain Research 1998), particularly on the right; with word generation tasks semantic and phonologic cues activate similar brain regions including anterior cingulate, left frontal cortex, thalamus and cerebellum, but subtle differences exist between them that are consistent with the lesion literature (Gourovitch et al, submitted). We have also carried out several experiments aimed at investigating the working memory system in normals under conditions that have relevance for cognitive symptoms of schizophrenia. First, in a dual-task paradigm that may model patients' limited working memory capacity, a fundamental characteristic of prefrontal cortex was activation was attenuated in the face of supramaximal demands for stimulus processing and response selection, at least when examined in the dual task condition (Goldberg et al, NeuroImage, 1998); these results may provide an analog for understanding cognitive failures in schizophrenia where both capacity limitations and reduced cerebral activation have been observed. Second, in experiments exploring the effects of increasing working memory load within working memory capacity, increasing the "working" (i.e. number of manipulations on remembered material) and the "memory" (i.e. the amount of material) within working memory produced markedly different patterns of activation within the frontal lobes. Increasing the amount of remembered material produced graded physiological responses primarily in areas that play a role in preparing a response within the motor domain, including verbal responses. Increasing the degree to which the remembered material was manipulated produced graded physiological responses in regions more classically associated with the working memory system, including dorsolateral prefrontal cortex. In a study of cognitive activation in normal aging we found that neurophysiological changes were context dependent (Esposito et al, submitted). That is, apparent changes over the life span differed in different tasks, depending on the role of the particular neural system for the particular task. In regions where physiological activity is normally suppressed when young subjects perform the tasks older subjects activate more, and, moreover, the more they activate (or fail to suppress), the worse they perform on the tasks. In other areas, where physiological activity is normally increased in young people performing the tasks, older subjects activate less; and the less they activate these regions, the more impaired their performance. We have developed a technique for estimating voxel-level statistical power associated with statistical parametric maps in cognitive activation studies and illustrated the effects of image smoothing on the expected reproducibility of significant activation (Van Horn et al, NeuroImage, 1998). Forty subjects were scanned during the WCS paradigm. The registered data were analyzed with SPM and voxel-wise statistical power estimates were generated for t-map result images employing the Non-Central F-Distribution method for measuring power. At a critical alpha of 0.01 and power thresholds of 80% and 95%, the power to reject the null hypothesis in brain regions implicated in the task at N's of 5 and 10 may not be sufficient to assure replicability of significant findings. When sample sizes exceed 20 subjects, power above 90% was found in the right DLPFC, bilateral inferior parietal lobule, and bilateral inferior temporal lobe, comprising the cortical network typically observed during the WCS. The filter size needed to maximize statistical power varied widely, but systematically, across regions, tending to follow known anatomical landmarks and and the juxtaposition of functionally reactive brain structures